What is the rPM value? Reads per million, maybe? If I'm reading this right, and these are relative frequency values of some kind, for three samples and 325 features, what you have is basically a FeatureTable[RelativeFrequency]. Does this sound right to you? The most common alpha diversity measures accept this semantic type, which means you may be in luck!
If this sounds like your situation, your mission would be to import your data as a FeatureTable[RelativeFrequency]. You could do this by reading the CSV into python and importing from a pandas dataframe, by converting the CSV into a TSV and then a biom.table, and importing that, or probably a number of other other nifty ways I'm overlooking. Does that help point you in the right direction?
Best,
Chris 